Uploaded January 2026 | Updated November 2026, 3 hours ago
Part 2 youtube.com/watch?v=MkrbecD-bK4
This is a push-pull approach with some common grounded IRF3205's aiming for 24Vin max and maybe 18Vout max, hopefully 10W or more if desired. Modern converters are much smaller than this pushing way more power but I don't really need it to be too small and thruhole is all I got anyway. The feedback involves driving an optocoupler with the high side output, and regulating the current it can sink through the TL431 which basically has a precise 2.5V internal reference pin. With some external current limiting resistors I can try to keep a stable supply output voltage by using this to trim around the led current which is managing the feedback through the collector output. I think the reason I had problems with Pin 3 feedback is because I wasn't sure what to do with error amp inputs. I tried using a 15K pullup to Vref but I can't remember if I left the error amps floating or just grounded them (yea lol I can see I grounded them). Maybe giving each one a 2.5V reference off the 5V ref with a couple dividers would have worked, but it seems fine just using the first amp and grounding the second one. At any rate the concept is probably most easily explained in that you're creating a setpoint where an action occurs, in this case a collector output pulldown and so on.
Could also use an auxiliary winding but I guess most decent isolated supplies use opto feedback.
Part 2 I will start looking into all the ways to improve the switching which not only involves the main power stage but all the collector outputs and how I'm driving those. Since I was using the Vref pin as a small current source at first I didn't want to pull it too hard so I probably have some slow switching going on before I even get to the UCC inputs. Couldn't figure any way to really deal with potential flux walk in this topology and played around the idea of the series cap and parallel resistor but that's probably more for low power converters. In that regard the half bridge would have probably been better but only really if I was looking to push higher voltages.
I currently could have some on time loss in UCC input relative to the TL494 collector active on time etc. Or more like the transformer is not suitable.
If everything looks good in the actual signal side I'll just have to keep playing with snubbers or redesign the power section to copper or something. I'm also just kind of lying to myself that I don't need to change the transformer cause I'm sure I do, and most definitely will.
Part 2 youtube.com/watch?v=MkrbecD-bK4
This is a push-pull approach with some common grounded IRF3205's aiming for 24Vin max and maybe 18Vout max, hopefully 10W or more if desired. Modern converters are much smaller than this pushing way more power but I don't really need it to be too small and thruhole is all I got anyway. The feedback involves driving an optocoupler with the high side output, and regulating the current it can sink through the TL431 which basically has a precise 2.5V internal reference pin. With some external current limiting resistors I can try to keep a stable supply output voltage by using this to trim around the led current which is managing the feedback through the collector output. I think the reason I had problems with Pin 3 feedback is because I wasn't sure what to do with error amp inputs. I tried using a 15K pullup to Vref but I can't remember if I left the error amps floating or just grounded them (yea lol I can see I grounded them). Maybe giving each one a 2.5V reference off the 5V ref with a couple dividers would have worked, but it seems fine just using the first amp and grounding the second one. At any rate the concept is probably most easily explained in that you're creating a setpoint where an action occurs, in this case a collector output pulldown and so on.
Could also use an auxiliary winding but I guess most decent isolated supplies use opto feedback.
Part 2 I will start looking into all the ways to improve the switching which not only involves the main power stage but all the collector outputs and how I'm driving those. Since I was using the Vref pin as a small current source at first I didn't want to pull it too hard so I probably have some slow switching going on before I even get to the UCC inputs. Couldn't figure any way to really deal with potential flux walk in this topology and played around the idea of the series cap and parallel resistor but that's probably more for low power converters. In that regard the half bridge would have probably been better but only really if I was looking to push higher voltages.
I currently could have some on time loss in UCC input relative to the TL494 collector active on time etc. Or more like the transformer is not suitable.
If everything looks good in the actual signal side I'll just have to keep playing with snubbers or redesign the power section to copper or something. I'm also just kind of lying to myself that I don't need to change the transformer cause I'm sure I do, and most definitely will.










